Tight-binding calculations of optical matrix elements for conductivity using non-orthogonal atomic orbitals: Anomalous Hall conductivity in bcc Fe

TL;DR

This paper derives a general formula for calculating optical matrix elements in tight-binding models with non-orthogonal atomic orbitals and applies it to compute the anomalous Hall conductivity in bcc Fe, showing good agreement with experiments.

Contribution

The paper introduces a universal formula for momentum matrix elements in tight-binding models with non-orthogonal orbitals, facilitating accurate conductivity calculations from first-principles.

Findings

The derived formula accurately reproduces the anomalous Hall conductivity in bcc Fe.

Introducing bandwidth renormalization improves agreement with experimental data.

The approach simplifies conductivity calculations in first-principles tight-binding models.

Abstract

We present a general formula for the tight-binding representation of momentum matrix elements needed for calculating the conductivity based on the Kubo-Greenwood formula using atomic orbitals, which are in general not orthogonal to other orbitals at different sites. In particular, the position matrix element is demonstrated to be important for delivering the exact momentum matrix element. This general formula, applicable to both orthonormal and non-orthonormal bases, solely needs the information of the position matrix elements and the ingredients that have already contained in the tight-binding representation. We then study the anomalous Hall conductivity in the standard example, ferromagnetic bcc Fe, by a first-principles tight-binding Hamiltonian. By assuming the commutation relation $\hat{\vec{p}} = (i m_e/\hbar) [\hat{H},\hat{\vec{r}}]$, the obtained frequency-dependent Hall conductivity is found to be in good agreement with existing theoretical and experimental results. Better agreement with experiments can be reached by introducing a reasonable bandwidth renormalization, evidencing the strong correlation among 3$d$ orbitals in bcc Fe. Since a tight-binding Hamiltonian can be straightforwardly obtained after finishing a first-principles calculation using atomic basis functions that are generated before the self-consistent calculation, the derived formula is particularly useful for those first-principles calculations.